Rats learned drug discriminations in a shock-escape T-maze task. They were trained to turn right in the maze following injection of a drug (D) and left when no injection (N) was given. Number of training sessions before criterion performance (STC) was used to indicate degree of discriminability of the training drug. STC decreased monotonically as dosage increased, and reached a minimum of 3 to 26 with various agonists. Most agonists were not highly discriminable. Daily maintenance injections of morphine, 200 to 600 mg/kg, increased the STC of morphine, 15 mg/kg, significantly, but complete tolerance to discriminable drug actions was not observed. After rats discriminated D vs. N, they were tested with novel drugs to determine which would elicit D choices. Most morphine-like agonists substituted for one another during substitution tests; the tested agonists included alphaprodine, codeine, fentanyl, heroin, meperidine, methadone, morphine, piminodine and propoxyphene. In a few instances, one of these agonists failed to substitute for another. Naloxone and naltrexone antagonized the discriminable effects of morphine. Cyclazocine, levallorphan, naltrexone, dextromethorphan, ethoheptazine and the narcotic agonists did not substitute for one another, suggesting that six dissimilar discriminable effects were produced by these drugs.